Key statistical indicators of beach litter survey results¶
Identification, quantification and analysis of observable anthropogenic debris along swiss river and lakes¶
The key indicators are common relationships used to provide insight to the most frequent questions using parameters that are taken directly from the survey data. The key indicators are relatively easy to calculate and interpret. In this section we explain the key indicators and use examples from the current data.
Indicators for the most frequent questions¶
- What do you find?
- How often do you find it?
- How much do you find?
- What else do you find?
- Where do you find the most?
These are important questions to answer. The survey results are a census of objects that were primarily washed up on the beach. This helps answer the question:
What are we likely to find at the beach?
What are we likely to find in the water?
The key indicators differ between regions and locations. Which may mean that the extent and/or nature of the problem is different from one region to another.
The reliability of these indicators is based on the following assumptions:
- The more trash there is on the ground the more a person is likely to find
- The survey results represent the minimum amount of trash at that site²
- For each survey: finding one item does not effect the chance of finding another³
More information ¶
For information on a specific catchment area or a water feature see the notebook for that catchment area (or make one and send a pull request). For more information on the project visit https://www.plagespropres.ch/ .
¹ The EU guide on monitoring marine litter https://mcc.jrc.ec.europa.eu/documents/201702074014.pdfhttps://mcc.jrc.ec.europa.eu/documents/201702074014.pdf
² There is most likely more trash at the survey site, but certainly not less than what was recorded.
³ Independent observations : https://stats.stackexchange.com/questions/116355/what-does-independent-observations-meanhttps://stats.stackexchange.com/questions/116355/what-does-independent-observations-mean
A brief reminder of the survey method¶
A survey is a collection of observations. The observations correspond to the objects that were removed and counted during the survey. Each object is placed into one of 260 categories¹. The location, date, survey dimensions and the total number of objects in each category is noted.
Some locations are sampled monthly, other were only sampled once.
The scope and size of the data for this example:¶
Biel/Bienne is the Aare survey area.
Fail rate: likelihood of finding an object¶
The fail rate is the number of times that an object was found divided by the number of surveys.¶
What does it mean? The fail rate describes the percent of times that a category was identified in relation to the amount of surveys conducted
Use the fail rate to determine how frequently an object is found within a geographic range
Use the fail rate to indentify objects that are found frequently but in low numbers
Use the fail rate and pcs/m to identify objects that are found infrequently but in important quantities
Calculate the fail rate per object category, for one city, lake and river basin¶
The fail rate can be calculated for any lake, municipality or river bassin provided you have a sufficient quantity of reliable data. Biel/Bienne is a city on Bielersee in the Aare river bassin. There were multiple surveys from multiple locations within the river bassin.
How much data is sufficient? The data should cover the land use characteristics for the geographic and temporal scale appropriate to the area of study. Local authorities will have insight to land use characteristics that can greatly increase monitoring effiency. How confident we are in the findings is directly related to the quality and the amount of data available.
To calculate the fail rates for Biel, Bielersee and the Aare river basin we just add up the number of times a code was used and divide it by the number of surveys for the city, lake or river basin.
The 10 items identified the most often in Biel/Bienne¶
Compare the fail-rates of the ten most common items from the 17 surveys in Biel to the fail-rates of those same items for Bielersee the Aare and all other survey areas.
With the exception of fragmented plastics and plastic sheeting the fail rate for the top ten items in Biel/Bienne was greater than the rest of the lake, the river bassin and nationally. This means that, in general, there was a greater chance of finding those objects at Biel/Bienne than most other places.
The fail rate is not the probability of finding one object, it is the most likely estimate (MLE) of the probability of finding one object. The MLE is the best estimate for the probability of a binomial variable (the pass fail rate is a bimomial variable). A complete derivation of the MLE of the binomial variable is beyond the scope of this article but very easy to understand⁴.
A 100% fail rate does not mean that you are guaranteed to find the object, it means that the object was identified in all previous samples.
With that the first two questions are answered:
Conclusion: We now know what items were found the most often and that tells us about what we may find at the next survey but we do not know how many of these objects were found at each survey. To do that we need to look at the quantity found as well as size of the survey.
⁴ A very simple explanation of the MLE for a binomial variable: https://www.youtube.com/watch?v=4KKV9yZCoM4
Pieces per meter: How many objects found per length of shoreline¶
Pieces per meter is the number of objects found at each survey divided by the length of the survey.¶
What does it mean? Pieces per meter describes the quantity of an object that was found for each meter of shoreline surveyed.
Use pieces per meter to find the objects that were found in the greatest quantities
Use pieces per meter to identify zones of accumulation
Why not use the surface area? The norm internationally is to report the results as quantity of objects per length of shoreline surveyed, usually 100 meters. You can use either one, however if you are looking for comparable data sets your choices may be limited if using the surface area is a requirement. The example here is given in pieces per meter.
Biel/Bienne: the ten objects with the highest median pieces per meter per survey¶
The pieces per meter (pcs_m) is calculated for each observation at the time the survey is submitted. This value can be taken directly from the survey results. The pcs_m value is not cumulative, therefore we need to use either the mean or the median value. Because we are interested in how many may be found at a single survey we will use the median value for each object at each survey in the city of Biel/Bienne.
The first thing to notice is that not all of the most frequently found objects are found in the greatest quantities.
With this observation we can answer questions 3 and 4: How much do you find? What else do you find?¶
By combining the average pieces per meter per survey of the results from above we are accounting for on average ~3.4 out of the 6 pieces per meter.
What about the rest? There were 121 different categories used, so far we have accounted for 2,279 of the 3,309 objects (68%) with only 14 categories. That leaves ~ 1,030 objects spread out between 107 categories⁵.
This result is not uncommon, a relatively small group of objects make up a large percentage of the objects found. This is true even in the marine environment a quick check of the survey results from Marine Litter Watch⁶ reveals that the top ten in Biel/Bienne share four common objects with the top ten items in the EU:
- Food wrappers; candy, snack packaging
- Foam polystyrene; expanded beads/balls 2.5cm - 50cm
- Cigarette butts and filters
- Plastic pieces 2.5cm - 50cm
⁵ Find the full list of objects found at the end of this article
⁶ https://www.eea.europa.eu/themes/water/europes-seas-and-coasts/assessments/marine-litterwatch/data-and-results/marine-litterwatch-data-viewer
Putting it all together¶
The pieces per meter and fail rate help define the trash removed with more precision. The fail rate for an item or a group of items is an indicator of how likely we are to find the item or group of items. Pieces per meter gives a reference value for the minimum number of objects we expect to find given the previous results. The more samples we have the more confident we can be in our assumptions.
This opens the possibility of measuring mitigation techniques and reduction strategies using both of these indicators. The pieces per meter and the fail rate can vary from one location to another and from one season to another. If both the fail rate and the pieces per meter decline for a group of objects or an object year over year, it is safe to assume that the abundance of the object in the environment is in decline.
Next steps¶
Understanding how the fail rate and pcs/m vary in response to things like population, infrastructure or time of the year.
Grouping codes together to represent economic sectors or usage.
Practical excercise 1: Where would you go to find plastic industrial pellets on Bielersee?¶
Plastic industrial pellets are the primary material used to produce plastic objects. They are disc or pellet shaped with a diameter of ~5mm.
Given the following survey results and the map of survey locations for Bielersee and assuming the surveys were done according to protocol:
- Where do you think your chances are the best of finding one?
- Suppose you did a survey of 50 meters. How many would you expect to find?
- Why did you pick that location? How sure are you of your choice?
Have a great day¶
This project was made possible by the Swiss federal office for the environment.
This document originates from https://github.com/hammerdirt-analyst/iqals all copyrights apply.